This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [v3] libstdc++/21726


On Thu, May 26, 2005 at 11:58:51AM +0200, Paolo Carlini wrote:
> Paolo Carlini wrote:
> 
> >2005-05-25  Paolo Carlini  <pcarlini@suse.de>
> >
> >	PR libstdc++/21726
> >	* configure.host (abi_baseline_pair switch): Add an entry
> >	for powerpc64-linux.
> >	* config/abi/powerpc64-linux-gnu/baseline_symbols.txt: New.
> >	* config/abi/powerpc64-linux-gnu/32/baseline_symbols.txt: New.
> >	* config/abi/powerpc64-linux-gnu/32/nof/baseline_symbols.txt: New.
> >  
> >
> Maybe a clarification is in order: the above configuration of
> directories is not going to work ok together with special -with-cpu=
> parameters. For instance, Janis is using for testresults,
> --with-cpu=default32, which *apparently* leads to an abi_check failure:
> in fact a 32-bit library (the default for that setup) is being compared
> to a 64-bit library.
> 
> I'm completely open to suggestion about a more general structure of
> directories, that would also work across --with-cpu= settings different
> from the default.

I run the testsuites with "--target_board=unix'{-m32,-m64}'", making the
other tests run with both -m32 and -m64.  I see abi_check being built
both ways and two summaries, but both summaries are identical.

Configuring GCC on powerpc64-linux to generate 32-bit objects by default
is done for the compilers in SLES9 and RHEL4, so it's not unusual.  See
config/rs6000/t-linux64 for the usual multilib directories.

Janis


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]